Justice DURHAM,
opinion of the Court:
INTRODUCTION
T 1 The Department of Workforce Services (DWS) denied Amy Sawyer's application for unemployment benefits based upon its finding that she quit her job without good cause. Ms. Sawyer appeals from the denial, arguing that (1) we should review DWS's good cause determination de novo and (2) DWS's good cause determination was in error. We conclude that good cause to quit is a fact-like mixed question of law and fact that we review deferentially. We also conclude, however, that the administrative law judge and appeals board applied an incorrect legal standard to this mixed question, and we therefore reverse and remand for further proceedings.
BACKGROUND
12 Ms. Sawyer was a special education teacher for the Jordan School District. The principal of the school where Ms. Sawyer worked became concerned with her teaching skills and informed Ms. Sawyer that she would be formally evaluated under the Jordan Performance Appraisal System (JPAS), which consisted of classroom observation by the principal and an interview. Ms. Sawyer received an overall score of "ineffective" during her first JPAS evaluation. After the failed evaluation, the principal provided suggestions for improvement and administered another JPAS evaluation the following month. Ms. Sawyer also received a failing score for her second JPAS evaluation.
T3 Following the second evaluation, the principal met with Ms. Sawyer and informed her that she would be required to complete a third JPAS evaluation. If Ms. Sawyer passed the third evaluation she would keep her job; but if she failed again, school policy dictated that she be terminated. The principal also told Ms. Sawyer that she could resign in order to avoid the third evaluation.
T4 Because she had some special-needs students with behavioral challenges in her class, Ms. Sawyer had little confidence that she could perform at a level that would allow her to pass a third evaluation. Ms. Sawyer was also concerned that if she were terminated that she would not find future employment as a teacher because schools typically ask whether an applicant has ever been fired from a teaching position. Therefore, Ms. Sawyer elected to resign rather than submit to a third JPAS evaluation.
15 Ms. Sawyer began searching for new employment and applied for unemployment benefits. DWS denied unemployment benefits because it found that Ms. Sawyer quit her job without good cause. An administrative law judge upheld the department's decision, reasoning that if Ms. Sawyer had chosen to submit to the third JPAS evaluation, "Islhe may not have lost her job." The Workforce Appeals Board affirmed the denial of benefits. The appeals board concluded that "[qluitting in order to avoid a discharge . does not establish good cause." Ms. Sawyer appealed, and the court of appeals certified the case to this court.
ANALYSIS
T6 An individual is ineligible for unemployment benefits if he or she quits "without good cause." Utax Cope § 85A-4-405(1)(a). We have adopted a reasonable person standard for determining whether good cause to quit exists: "Good cause is established where 'the unemployment is caused by pressures so compelling that a reasonably prudent person would be justified in quitting under similar cireumstances." Hurst v. Indus. Comm'n, 723 P.2d 416, 419 (Utah 1986); accord Smith v. Indus. Comm'n, 714 P.2d 1154, 1155 (Utah 1986). DWS has adopted rules in accord with the reasonable person standard, explain
ing that in order for an individual to qualify for benefits, "[tlhe separation must have been motivated by circumstances that made the continuance of the employment a hardship or matter of concern, sufficiently adverse to a reasonable person so as to outweigh the benefits of remaining employed." Utah Admin. Code R994-405-102(1)(a).
T7 In reviewing DWS's determination that Ms. Sawyer quit without good cause, we first establish the appropriate standard of review for this mixed question of law and fact.
See Smith, 714 P.2d at 1155 (good-cause determination is a mixed question). We then determine whether DWS erred when it denied benefits.
I. STANDARD OF REVIEW
A. Mixed Questions of Law and Fact
8 Prior to the formation of the court of appeals, this court did not consistently articulate the precise standard of review it applied to the cases before it. State v. Thurman, 846 P.2d 1256, 1268, 1270 n. 11 (Utah 1993). The addition of an intermediate court of appeals, however, created a greater incentive to establish standards of review that could be uniformly applied by both appellate courts. Id. Moreover, established standards of review allow attorneys to better advise clients on appellate matters and facilitate an appellant's ability to select and properly frame arguments.
T9 In line with our efforts to better define the standard of review applied by appellate courts, we created an analytical framework for choosing the standard of review for mixed questions of law and fact in State v. Pena, 869 P.2d 982 (Utah 1994). In Pena, we held that the amount of discretion afforded in an appeal from a district court's application of a rule of law to a given set of facts is an institutional policy determination made by the appellate court.
Id. at 938-39. Depending on the nature of the legal question at issue, we determined that varying levels of deference should be afforded to a district court's resolution of mixed questions. Id. at 937-38. We envisioned multiple standards of review for various types of mixed questions, occupying a spectrum of deference falling between the nondeferential de novo standard of review and the highly deferential clearly erroneous standard of review:
[The amount of deference] permitted a trial judge will vary depending on the legal issue, although the terminology we use to describe the operative standard of review does not begin to reflect the many shades of this variance. The best we can do is to recognize that such a spectrum of disceretion exists and that the closeness of appellate review of the application of law to fact actually runs the entire length of this spectrum.
Id. at 988.
{10 Although a broad spectrum of standards of review affords appellate courts a great amount of flexibility, the principal drawback of this approach is that it leads to indefinite standards that are difficult to describe and even more difficult for litigants or appellate courts to predict and apply. In Pena, for example, we held that the mixed question at issue in that case "conveys a measure of discretion to the trial judge," but "[plrecisely how much discretion we cannot say." Id. at 939. The Pena court went on to describe this standard of review as "something less than de novo" and acknowledged that "this 'some discretion' standard is less than precise." Id. at 940 & n. 6. Applying the principles announced in Pena, we have arrived at similarly indefinite standards of review, such as "some serutiny," "limited deference," and "conditionally deferential." Drake v. Indus. Comm'n, 939 P.2d 177, 182 (Utah 1997); State v. Virgin, 2006 UT 29, ¶ 34, 137 P.3d 787; Salt Lake City Corp. v.
Labor Comm'n, 2007 UT 4, ¶ 15, 153 P.3d 179.
111 While we have not rejected the broad spectrum approach espoused in Pena or the various standards of review adopted in subsequent cases, in our more recent cases we have applied a binary method for determining the appropriate standard of review for mixed questions,. In Manzanares v. Byington (In re Adoption of Baby B.), 2012 UT 35, ¶¶ 42, 44, 308 P.3d 382, we stated that mixed questions can either be law-like or fact-like. See also Jex v. Utah Labor Comm'n, 2013 UT 40, ¶ 15, 306 P.3d 799; Union Pac. R.R. v. Utah Dep't of Transp., 2013 UT 39, ¶ 15, 310 P.3d 1204; Murray v. Utah Labor Comm'n, 2013 UT 38, ¶ 37, 308 P.3d 461; Swallow v. Jessop (In re United Effort Plan Trust ), 2013 UT 5, ¶ 19, 296 P.3d 742. Law-like mixed questions are reviewed de novo, while fact-like mixed questions are reviewed deferentially. Baby B., 2012 UT 35, ¶ 42, 308 P.3d 382. We took this approach in another case involving unemployment benefits, Carbon County v. Workforce Appeals Board, 2013 UT 41, ¶ 7, 308 P.3d 477, and similarly conclude that the standard of review for the mixed question at issue here turns on whether it is properly characterized as either law-like or fact-like.
112 In determining: whether a mixed question should be deemed law-like or fact-like, we evaluate the "marginal costs and benefits" of conducting either a searching de novo review or a deferential review of a lower tribunal's resolution of the mixed question. Baby B., 2012 UT 35, ¶ 42, 308 P.3d 382. This cost-benefit analysis is conducted through the three-factor Levin evaluation, in which we consider
(1) the degree of variety and complexity in the facts to which the legal rule is to be applied; (2) the degree to which a trial court's application of the legal rule relies on facts observed by the trial judge, such as a witness's appearance and demeanor, relevant to the application of the law that cannot be adequately reflected in the record available to appellate courts; and (3) other policy reasons that weigh for or against granting discretion to trial courts.
State v. Levin, 2006 UT 50, ¶ 25, 144 P.3d 1096 (internal quotation marks omitted).
113 The first and second Levin factors assess whether a particular mixed question is best resolved by either a fact-finding tribunal or an appellate court based on the relative competencies of these two types of courts. District courts and fact finding administrative bodies are in a superi- or position to weigh facts that depend upon credibility determinations, the direct observation of witness testimony, and other evidence not fully captured in a written appellate record. The degree to which a mixed question is based upon facts observed by a lower tribunal determines whether the second factor weighs for or against a deferential standard of review. Appellate courts, on the other hand, have the capacity to create broad rules that can create a greater degree of consistency and predictability to future cases involving a particular mixed question. The degree to which a mixed question is based upon a complex variety of facts determines whether an appellate court can create useful precedent and, thus, whether the first factor weighs for or against de novo review.
114 The third Levin factor is a catchall category under which an appellate court may weigh other considerations. A prime example of these other policy reasons for reviewing mixed questions de novo can be found in our search and seizure jurisprudence. Despite the fact-intensive nature of determining whether police officers had reasonable suspicion to conduct a search or whether a warrant was supported by probable cause, we review these questions de novo in order to provide guidance to law enforcement officials:
We have not retreated ... from exercising de novo review of equally fact-intensive matters that arise in the realm of search and seizure. We have not ceded deference in these cases both because they concern constitutional rights and because irrespective of the difficulties inherent in extracting general rules from fact-intensive matters, we nevertheless believe it a worthy endeavor to offer some guidance to law enforcement officials charged with the duty of conducting their affairs within constitutional bounds.
Salt Lake City Corp., 2007 UT 4, ¶ 15 n. 1, 153 P.3d 179.
The mixed question of whether a defamatory statement was made with actual malice and the issue of whether speech may be punished as obscene are likewise reviewed de novo. Jensen v. Sawyers, 2005 UT 81, ¶¶ 91-92, 130 P.3d 325; City of St. George v. Turner, 860 P.2d 929, 932-33 (Utah 1998). Other mixed questions with constitutional dimensions that we have reviewed de novo for policy reasons include whether a police interrogation was custodial, Levin, 2006 UT 50, ¶¶ 41-42, 144 P.3d 1096, and whether a confession was voluntary, Thurman, 846 P.2d at 1271.
B. Good-Cause-to-Quit Determinations Are Fact-Like
4 15 In determining whether a lower tribunal's good-cause-to-quit determination is either law-like or fact-like, we first look to prior cases in which we have articulated a standard of review for this question. Because unemployment benefit cases are typically handled by the court of appeals, we have not decided a good-cause-to-quit case since developing a framework and terminology for determining the standard of review for mixed questions in Peng, Levin, and Baby B. Cases that predate Pena, however, are still relevant. Pena and its progeny "are not fundamental departures from earlier standard-of-review law. Rather, they clarify and further define basic positions that have long served as the foundation for standard-of-review law both nationally and within Utah." State v. Vincent, 883 P.2d 278, 281 (Utah 1994).
16 Although we have not used uniform terminology to describe the standard of review for a good-cause-to-quit determination, we have consistently stated that we cede a great deal of deference to DWS's resolution of this mixed question. Hurst v. Indus. Comm'n, 723 P.2d 416, 419 (Utah 1986) ("In reviewing ... what constitutes good cause [to quit] ..., we are bound to defer to [DWS] so long as its decision falls within the limits of reasonableness and rationality."); Smith v. Indus. Comm'n, 714 P.2d 1154, 1155 (Utah 1986) (same); Gibson v. Indus. Comm'n, 707 P.2d 675, 676 (Utah 1985) ("The definition of good cause requires the application of a legal question to the fact situation at issue. On questions of mixed law and fact, we will not substitute our judgment for that of [DWS] so long as its interpretation has warrant in the record and a reasonable basis in the law."); Box Elder Cnty. v. Indus. Comm'n, 632 P.2d 839, 841 (Utah 1981) ("[Good-cause-to-quit determinations], onee found to be supported by the evidence, are binding and will not be disturbed."). In addition, we have recently decided that a similar mixed question, whether an employer terminated an employee for "just cause," is fact-like in nature. Carbon Cnty. 2013 UT 41, ¶¶ 5, 7, 308 P.3d 477.
17 An application of the Levin factors also points toward a deferential standard of review. Under the first factor, the variety and complexity of factual seenarios that might ecause a worker to quit "does not lend itself to consistent resolution by a uniform body of appellate precedent." Baby B., 2012 UT 35, ¶ 42, 308 P.3d 382. Under the second factor, DWS's decision to grant or deny benefits will often rely on facts observed by an administrative law judge. Workers will inevitably be required to explain the reasonableness of their decision to quit based upon job conditions or other external considerations, and administrative judges must assess the
credibility of the worker or other witnesses to determine the degree to which this testimony is accurate or exaggerated. Thus, administrative law judges will have a more nuanced view of the facts than an appellate court would, and administrative judges will be in a better position to apply the good-cause-to-quit legal standard to these facts. Finally, under the third Levin factor, we do not find any other considerations that weigh in favor of de novo review.
1 18 Citing Murray and Baby B., Ms. Sawyer contends, however, that good-cause-to-quit determinations should be reviewed de novo in a manner similar to probable cause or reasonable suspicion determinations under our search and seizure jurisprudence. Ms. Sawyer argues that because both search and seizure cases and good-cause-to-quit cases turn upon the general reasonableness of the action taken by the police officer in conducting a search or a worker in deciding to quit, both types of mixed questions are law-like. In support of her argument, she cites Baby B., where we state:
[The downside [of reviewing probable cause and reasonable suspicion cases de novo) is minimal in a case involving common, recurring practices, where the decision will turn on the general reasonableness of those practices and not so much on the demeanor or credibility of a particular witness. This is why a mixed finding of reasonableness is typically subject to a non-deferential standard of review.
2012 UT 35, ¶ 44, 308 P.3d 382 (footnote omitted); accord Murray, 2013 UT 38, ¶ 39, 308 P.3d 461 (a search and seizure determination "is not 'fact-like' because the ultimate determination will often rest on the 'general reasonableness' of the facts").
19 Ms. Sawyer's interpretation of these cases is incorrect, Baby B.'s reasoning did not depend on a simple rule that all determinations of reasonableness are law-like. Rather, Baby B. used the Levin factors to analyze the costs and benefits of de novo review for each of the mixed questions it had to address. 2012 UT 35, ¶¶ 42-46, 308 P.3d 382.
T 20 This analysis led the Baby B. court to conclude that some determinations of reasonableness should be reviewed de novo and others should not. On the one hand, a finding that a search or seizure was reasonable is subject to de novo review in part because search and seizure cases involve "common, recurring practices" that allow appellate courts to establish "a consistent rule ... as to the reasonableness of certain law enforcement procedures." Id. 144. Thus, "a mixed finding of reasonableness is typically subject to a non-deferential standard of review" in search and seizure cases. Id.
121 In other contexts, mixed findings of reasonableness are more often not subject to non-deferential review, and Baby B. mentioned two of these. In negligence suits, the reasonableness of a defendant's conduct is determined by the fact-finder and subject only to deferential review. See id. ¶ 43. And in Baby B. itself, we gave "some deference" to a district court's determination that a biological father could have known, through "the exercise of reasonable diligence," that his child might be born in Utah. Id. ¶ 46.
122 In both of these contexts, appellate courts' ability to lay down clear rules is minimal because of the "complex and varying" cireumstances to which the rules would have to apply. Id. (internal quotation marks omitted). It would be impossible for appellate courts to "spell[ ] out," in advance, how each potentially relevant fact should affect the legal outcome, and we have therefore entrusted these determinations of reasonableness to district judges' superior knowledge of the evidence of the case and to trial juries' superior knowledge of the community standards that govern the behavior of reasonable people. Id. (internal quotation marks omitted).
1 23 Ms. Sawyer also argues that this court should review the administrative decision below de novo because the facts are not in dispute. But this argument also misconstrues our precedent. In Murray, 2013 UT 38, ¶ 40, 308 P.3d 461, we indicated that the second Levin factor may weigh in favor of de novo review where the facts are not at issue because the lower tribunal is not required to evaluate witness credibility or demeanor. But the other two Levin factors may weigh
more heavily in favor of deferential review of a given mixed question. For example, if a plaintiff and a defendant agreed to the essential facts of a negligence claim, leaving only the mixed question of whether the defendant acted negligently given those facts, we would still not review a jury's resolution of this mixed question de novo, for the reasons explained above. See supra ¶¶ 21-22.
But even if our precedents did support the notion that de novo review is appropriate wherever the facts are not in dispute, that would not help Ms. Sawyer here. In this case, contrary to her assertions, important facts were in dispute. Ms. Sawyer's central contention is that she quit for good cause because, despite her reasonable efforts, the principal of her school would have given her a failing mark if she had participated in a third performance evaluation, leading to her termination. The school district did not concede that Ms. Sawyer's prediction that her termination was inevitable was accurate. Instead, the principal of her school testified that if Ms. Sawyer's performance had improved in the third evaluation, she would not have been terminated. Because the likelihood that Ms. Sawyer would have passed the third evaluation was disputed and involves a credibility determination of testimony produced by the school district, the second Lev-in factor supports deferential review.
125 Considering the relevant precedent and the Levin factors, we conclude that a good-cause-to-quit determination is a fact-like mixed question, and we apply a deferential standard of review to a lower tribunal's resolution of this issue. However, "we must be vigilant in our review of ... mixed findings to ensure that they are based on correct legal principles." Baby B., 2012 UT 35, ¶ 47, 308 P.3d 382. We review the legal standard applied to a particular mixed question for correctness. Id.
IL GOOD CAUSE TO QUIT
126 In evaluating Ms. Sawyer's contention that DWS erred by finding that she quit without good cause, we first evaluate whether it applied the correct legal standard. If a lower court or administrative body does not apply the correct legal standard to a mixed question, we must reverse. See Price River Coal Co. v. Indus. Comm'n, 731 P.2d 1079, 1083 (Utah 1986).
127 Ms. Sawyer consistently asserted in the proceedings below why she quit her job. She believed that, despite her reasonable efforts, she would not pass the third teaching evaluation and her resulting termination would prevent her from finding another teaching job.
The administrative law judge that evaluated her claim found that "[tlaking the third JPAS evaluation was a reasonable alternative to quitting and [Ms. Sawyer] did not take it. She may not have lost her job if she had." In other words, the administrative judge concluded that the possibility that Ms. Sawyer could have passed the third evaluation dictated that good cause to quit was not established. The Workforce Appeals Board affirmed the legal conclusion drawn by the administrative judge, stating unequivocally that "[qluiting in order to avoid a discharge ... does not establish good cause."
$28 Utah courts have not considered whether quitting in order to avoid a potential discharge may ever constitute good cause. The Oregon Supreme Court, however, has considered this question in McDowell v. Employment Department, 348 Or. 605, 236 P.3d 722 (2010). In that case, a school district informed a high school teacher that it would recommend the teacher's termination at an upcoming school board meeting for showing a film clip containing profane language to his students. Id. at 724. The teacher's union attorney advised him to quit because there was no chance that the school board would overrule the school district's recommendation. Id. Fearing he would be unable to find another teaching job if he were terminated,
the teacher resigned. Id. at 724, 727-28. Oregon's employment department denied unemployment benefits, and an administrative law judge and an appeals board affirmed the denial,. Id. at 724-25. The appeals board concluded that because the teacher faced "a mere possibility of discharge," and because his termination "was not a foregone conclusion," he did not have good cause to quit. Id. at 728 (internal quotation marks omitted).
11 29 The Oregon Supreme Court reversed, reasoning that
[tlo the extent that the board has suggested ... that a future discharge must be "certain" before a resignation to avoid the discharge can qualify as good cause, such a conclusion would be inconsistent with the "reasonable and prudent person" standard in place.... [The fact that a threatened discharge is less than certain to occur ... [is not] dispositive. ...
Id. at 780 n. 9. Instead, the court concluded that good cause to quit depends "on whether a reasonable person facing that prospect of discharge would consider the prospect so grave a cireumstance that resigning was the only reasonable option." Id. at 730; see also Madisonville Consol. Indep. Sch. Dist. v. Tex. Emp't Comm'n, 821 SW.2d 310, 318 (Tex.Ct.App.1991) (a teacher may quit with good cause if the teacher has "good reason to believe that he will imminently be discharged ... unless he chooses to resign"). This "objective inquiry depends on what [the] claimant in fact knew and reasonably should have known when he made his decision, not on an assessment of how events in fact would have played out." McDowell, 236 P.3d at 730.
$380 We similarly hold that the administrative judge and appeals board erred by concluding that the possibility that Ms. Sawyer could have retained her job was suffi-clent to defeat her employment benefits claim. Good cause to quit is measured by the objective standard of whether "a reasonably prudent person would be justified in quitting under similar cireumstances." Hurst v. Indus. Comm'n, 723 P.2d 416, 419 (Utah 1986). This assessment should be based on the information that the worker knew or should have known at the time of the resignation. McDowell, 236 P.3d at 730. Reasonably prudent persons, of course, must often make decisions based upon an assessment of potential consequences rather than in the context of certain outcomes. Little in life is guaranteed. In evaluating whether a reasonably prudent employee would quit in order to avoid a potential termination, administrative law judges and courts should consider (1) the likelihood of termination, in spite of the employee's reasonable efforts to remain employed, and (2) the degree to which termination will negatively affect future employment.
CONCLUSION
31 Because the administrative law judge and appeals board did not assess whether a reasonable person in Ms. Sawyer's shoes would have quit, but rather whether there was some possibility that she could have retained her job, we conclude that an incor- ' rect legal standard was applied to the facts of this case. We therefore reverse and remand Ms. Sawyer's unemployment benefits claim to the administrative law judge for further proceedings consistent with this opinion.